A new voice AI model claims to run on budget Android phones without an internet connection or per-query fees. VoxNexus, a Singapore-based company, announced the release of VoxEdge on October 23, 2026, through the APO Group in Madrid, Spain.
VoxEdge is an on-device voice stack for speech recognition, translation, and speech synthesis. It supports speech recognition in 36 languages, translation in 103 languages, and speech synthesis in 45 languages. The release highlights Swahili, Hausa, and Yorùbá but does not specify which African languages are supported for each function.
Once installed, the model operates entirely on the device, ensuring free queries without cloud interaction. VoxNexus also introduced VoxSense, an always-on listening tool that consumes 162 mAh over 12 hours, approximately 2.5% of a 6,500 mAh battery. Both tools are available for licensing through an SDK, with a free trial option.
The company claims its speech recognition model has 65 million parameters and outperforms larger open-source models in word error rate. However, the announcement lacks benchmarks, test sets, or comparison models for verification. Charles Shi, CEO of VoxNexus, emphasized the importance of under-resourced languages, noting that hundreds of millions of people speak them.
VoxNexus demonstrated VoxEdge at the MediaTek Dimensity 9600 launch in Madrid on October 19, 2026. The live demo transcribed Spanish and English conversations in real time without cloud connectivity. Notably, the African languages mentioned in the release were not part of this demonstration.
The announcement omits details on compatible phones, chipsets, pricing, or African customers. It also does not provide accuracy figures for individual languages or confirm the performance of African languages compared to demonstrated ones.
Limited data has historically constrained speech technology for African languages. A Google Research and Zindi challenge in September 2026 highlighted the gaps in existing systems. Additionally, AethexAI secured $3 million in funding for voice AI tailored to low-bandwidth African networks. VoxNexus positions its offline, on-device processing as a solution to both data and connectivity challenges.
The release promises broad language support, but the absence of specific details and performance metrics leaves room for further scrutiny. VoxNexus’s focus on under-resourced languages aligns with growing efforts to expand AI accessibility globally.
